JPWO2013179335A1 - Surveillance camera control device and video surveillance system - Google Patents

Abstract

A monitoring camera control device and a video monitoring system for selecting an appropriate video from video for each camera when the same object is detected redundantly in a plurality of monitoring cameras that capture a plurality of areas. An object is detected in a plurality of cameras 100 to 102 that image the inside of the monitoring area, a recognition unit 104 that detects an object from images acquired by the plurality of cameras, and a monitoring area that is imaged redundantly by the plurality of cameras. In this case, a recognition result that is a feature amount of an object for each camera is acquired in the recognition unit, and a display selection that gives priority to the video for each camera according to the priority based on the recognition result and the priority of the recognition result. The unit 106 is provided.

Description

  The present invention is a video surveillance system that has a recognition device for detecting a person or a moving body from video acquired from an imaging device such as a camera, and realizes functions such as intruder detection and approaching person detection installed in a mobile robot. In particular, the present invention relates to a video surveillance system having video acquisition and video display control functions of the video surveillance system.

  There is a video monitoring system having a function of performing recognition processing on video acquired from an imaging device such as a camera and detecting a moving object such as a person or a vehicle appearing in a monitoring area. This video monitoring system has a function of recording only a video in which a moving object appears by using a detection result, a function of presenting a warning icon on a display device, a function of sounding a buzzer or the like to alert a monitor. For this reason, it has helped to reduce the burden of monitoring work that previously required constant confirmation work. In addition, in this video surveillance system, when a criminal act such as theft or an illegal act occurs, the recorded video can be used for a subsequent criminal investigation.

  In recent years, crime diversification, an increase in the number of crimes, and a decrease in the clearance rate have heightened awareness of crime prevention at mass retailers, financial institutions, buildings and offices, and video surveillance systems have been introduced. Video recording devices are also increasing in capacity, and the number of cameras is increasing as cameras are installed in various locations due to the spread of network cameras and the like. As described above, since it is extremely burdensome to identify criminal acts from recorded images by visual observation by the observer, there is an increasing demand for a function that supports monitoring work.

  Therefore, the problem is that the increase in the number of cameras makes the task of observing a desired image, for example, a specific person, very complicated by a monitor. Unless it is an expert who knows the situation of a surveillance area and a surveillance camera, it is difficult to observe an image efficiently without overlooking it.

  Conventionally, a monitoring system has been known in which a moving position of a camera having a pan head control function is registered in advance, and an image of a monitoring area is acquired while turning the camera pan head to a predetermined position in conjunction with a detection result of the sensor. (Patent Document 1). There is also known a monitoring system in which positional relationships among a plurality of cameras are registered, and an image displayed on a monitor is switched based on a user instruction or a moving direction of a moving object (Patent Document 2).

JP 2001-275104 A JP 2011-217320 A

  However, in the monitoring system described in Patent Document 1, as the number of cameras operated in the system increases, the number of cameras turning in response to sensor detection also increases, so detailed settings and adjustments in advance in the system There is a tendency to increase the burden on observers.

  In addition, a time lag or the like occurs when the monitor control function is operated while being operated by a monitor, so that there is a high possibility that an intruder or a predetermined person will be overlooked.

  Therefore, there is a conventional system that has a recognition function and automatically tracks. For example, Patent Document 2 is known, and a positional relationship is registered in advance using a plurality of cameras installed in a monitoring area as fixed cameras, and the positional relationship is used as in the monitoring system of the present invention. is there. Specifically, when a moving object deviates from the angle of view on the screen of each camera, the camera installed in the moving direction is presented to the monitor. This can also be realized by designating the direction by the input means of the monitor.

  However, depending on the positional relationship of the cameras, it is possible to present the next camera image to be taken in consideration of the movement of the moving object, but depending on the installation status of the camera, it is better to display the image as it is There is. This is because tracking is performed based on the movement of a moving object as a subject, and is not suitable for obtaining information for identifying the subject (for example, a face for a person or a license plate for a vehicle or an area where a driver is not imaged) This is because it may be difficult to utilize for the situation evidence because the image is taken from the direction in which important information of the movable object cannot be obtained.

  In addition, when there are overlapping cameras, or when there is a degree of freedom in the movable area of the moving object, such as in an open surveillance area such as an office or retail store, rather than a surveillance area with a limited movable area such as a corridor As described in Patent Document 2, it is difficult to simply register the camera arrangement relationship in advance. This is because, for example, since the moving object moves to the right side of the angle of view, it does not have a simple structure of presenting the video of the camera installed on the right side, so the selection conditions for the camera that presents the video may be very complicated. It is done.

  Therefore, the present invention provides a monitoring camera control device and a video monitoring device that select an appropriate video from video for each camera when the same object is detected in duplicate in a plurality of monitoring cameras that capture a plurality of areas. Provide a system.

  In order to solve the above-described object, for example, the configuration described in the claims is adopted. For example, in a monitoring area that includes a plurality of cameras that capture images in a monitoring area and a recognition unit that detects objects from images acquired by the plurality of cameras, When an object is detected, a recognition result that is a feature amount of the object for each camera is acquired in the recognition unit, and priority is given to the video for each camera based on the recognition result and the priority of the recognition result. A display selection unit for ranking is provided.

  According to the present invention, in a plurality of surveillance cameras that capture a plurality of areas, when the same object is detected in duplicate, information useful for identifying an object or the like can be presented from a video for each camera. it can.

1 is a block diagram illustrating a video surveillance system according to an embodiment of the present invention. It is the figure shown about the arrangement | positioning of the monitoring area and camera of this invention. It is a bird's-eye view of a surveillance area and camera arrangement of the present invention. It is an example of an output of a picture acquired by the present invention. It is the figure shown about the correspondence of the camera image of this invention, and a monitoring area. It is the figure which showed the processing flow in the recognition part of this invention. It is the figure which showed the data structure of the recognition result of this invention. It is the figure which showed the data structure of the camera arrangement | positioning information of this invention. It is the bird's-eye view which showed the information of the camera and the moving object on the monitoring area by this invention. It is a figure which shows the information table calculated from the result acquired in the recognition part of this invention, and camera arrangement | positioning information. It is the figure shown about the priority setting means of this invention. It is the figure shown about the video display method of this invention. It is the figure shown about the video display method of this invention. It is the figure shown about the switching method of the video display of this invention. It is the block diagram shown about one Embodiment of this invention. It is the block diagram shown about one Embodiment of this invention.

  Hereinafter, an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

  FIG. 1 is a block diagram showing a surveillance camera control device and a video surveillance system according to an embodiment of the present invention. In this embodiment, the monitoring camera control device and the video monitoring system may be simply described as a video monitoring system or a monitoring system.

  This video surveillance system includes cameras 100 to 102, a video acquisition unit 103, a recognition unit 104, a recognition result 105, a display selection unit 106, camera arrangement information 107, an input unit 108, and a video display unit 109. The display means 110 is provided.

  This video surveillance system has a configuration to which an electronic computer system is applied. The hardware of this electronic computer system includes a CPU, a memory, an I / O, and the like, and each functional unit represented by a block in each figure is realized by installing predetermined software in an executable manner.

  In this example, the cameras 100 to 102 are described as three cameras in order to easily represent the embodiment, but the configuration is assumed to be independent of this configuration and two or more cameras are installed. . Each of the cameras 100 to 102 is an imaging device including a camera lens having a zoom function and an imaging element (none of which is shown) such as a complementary metal oxide semiconductor (CMOS) or a charge coupled device (CCD). The cameras 100 to 102 acquire a video signal by the video acquisition unit 103 and output the video signal to a recognition unit 104 and a video display unit 109 described later.

  The cameras 100 to 102 are pan / tilt / zoom cameras that are placed on a pan / tilt head and can be turned up and down. Although not described in the present embodiment, it is obvious that the images of the cameras 100 to 102 are transferred to a recording device or a display device, and the images are recorded or used for visual confirmation by a monitor.

  The display means 110 is a display device such as a liquid crystal display device or a CRT (Cathode Ray Tube) display device. Instead of providing the display unit 110, a configuration using RGB (Red-Green-Blue) monitor output, data output via a network, or a terminal such as a mobile phone or a tablet may be used.

  Various parameter settings are performed from the user interface. The user interface provided in the video acquisition unit 103, the recognition unit 104, the video display unit 109, and the like includes an input device (not shown) such as a mouse and a keyboard, and accepts input of parameters and the like by the user. For the description of the basic part of the present invention, only the input means 108 is described as means for inputting parameters and the like to the display selection unit 106.

  Next, before describing the details of the constituent blocks of the present invention, the relationship between the camera and the moving object in the monitoring system of the present invention will be described with reference to FIG.

  FIG. 2 shows the relationship between the monitoring area 205 of this embodiment and the cameras and objects installed there. In the monitoring area 205, cameras 200 to 202 (similar to the cameras 100 to 102 in FIG. 1) are installed, and an object 203 is present. In addition, furniture such as shelves and building structures such as walls and hallways often exist in the monitoring area 205 and are illustrated as structures 204. In this embodiment, assuming that the object 203 is a person, the object 203 is moving in the moving direction 206 and facing the face direction 207.

  Here, the object includes a movable object and a stationary object. A movable object refers to an object that can move and fluctuate. In the present embodiment, a person is taken as an example of a movable object, but a person is an object that can move and fluctuate as a face, limbs or the whole person. In addition, examples of the movable object include a vehicle, a bag carried by a person, a personal computer screen, a safe door, and the like. For example, the screen of a personal computer, the door of a safe, and the like are objects that can change depending on the person, such as the orientation of the screen, the screen display, and the opening of the safe door. The present invention can also be applied to a stationary object that does not move or fluctuate.

  The monitoring area 205 is used synonymously with real space and the like, and its coordinate system is defined in advance as (Xw, Yw, Zw).

  Next, FIG. 3 shows a bird's eye view when the monitoring area 205 is observed from above.

  Here, imaging regions 300 to 302 (used synonymously with angle of view) corresponding to each camera are newly shown. Others are the same as in FIG. 2, and the object 203 is present on the monitoring area 205, moves in the moving direction 206, and faces the face direction 207.

  FIG. 4 illustrates images captured by the cameras 200 to 202. Camera images 400 to 402 represent videos captured by the cameras 200 to 202, respectively, which are acquired through the video acquisition unit 103 in FIG. 1 and displayed on the display unit 110 through the video display unit 109. Images are captured including the object 203 and the structure 204 depending on the installation status of each camera and the imaging regions 300 to 302. Depending on the positional relationship with the cameras 200 to 202, the appearance of the object 203 and the structure 204, the size of the object 203, and the appearance of the moving direction 206 and the face direction 207 are different.

  In these drawings showing the monitoring area 205 and the object 203 described in FIGS. 2 to 4, the object 203 is moving in the direction of the camera 200, and the face direction 207 is facing the direction of the camera 201. . The object 203 is located on the lower right side from the center of the Xw-Yw space of the monitoring area 205, and the camera 202 is the closest camera.

  Here, an example of calculating the correspondence between the camera and the monitoring area will be described.

  In order to calculate the correspondence between the camera and the monitoring area, that is, the camera parameter, the present invention is not limited to this embodiment, and there are a method for obtaining the parameter in a simple manner and a method for obtaining it in detail. This correspondence is used for acquiring the camera arrangement information 107 shown in FIG.

  FIG. 5 is a diagram illustrating corresponding points on the camera image acquired by the monitoring area and the camera.

  Specifically, as shown in FIG. 5, a method of taking corresponding points on the monitoring area 501 (monitoring area 205, synonymous with real space) and the camera image 502 acquired by the camera 500 can be considered. The camera 500 exists at a certain position (Xc, Yc, Zc) on the monitoring area 501.

The correspondence between the arbitrary camera image position 504 on the camera image 502 and the monitoring area position 505 on the monitoring area 501 can be obtained from the position on the image and the actual measurement value in the real space. After acquiring the corresponding points, as a method of acquiring camera parameters, for example, “RY Tsai,“ A versatile camera calibration technique for high-accuracy 3D machine vision
It is known as an existing technology for camera calibration technology such as "metrology using off-the-shelf TV camera and lenses" IEEE Journal of Robotics and Automation, Vol.RA-3, No.4, pp.323-344, 1987 ". However, detailed description is omitted here. It is known that the camera parameter can be obtained by obtaining four or more points in the method for obtaining the camera parameter from the corresponding points.

  By this procedure, the depression angle θ of the camera 500, the installation angle φ on the monitoring area 501 and the camera height Hc can be obtained as shown in FIG.

  Next, description will be made in order from the recognition unit 104 shown in FIG. The recognizing unit 104 executes recognition processing on an arbitrary video among the plurality of videos acquired by the video acquiring unit 103.

  FIG. 6 is an example of a flowchart for explaining processing of the recognition unit 104, and illustrates detection of a human face and a face direction in the present embodiment.

  A method for detecting a human face has been widely proposed. For example, “Paul. Viola, M. Jones:“ Robust Real-Time Face Detection ”, International Journal of Computer Vision (2004), Volume: 57, Issue: 2 , Publisher: Springer, Pages: 137-154. They obtain facial image features from learning samples and construct a classifier. This discriminator determines where the face is on the image. In addition, it is possible to recognize the face direction by constructing each classifier after dividing the learning sample into various partial samples such as front and side. Hereinafter, it will be described with reference to FIG.

  S60 is a procedure for operating the entire image in an arbitrary window (detection window). Thereafter, whether or not a face has been detected using the above-described discriminator is output at a certain arbitrary position (S61). If no face is detected, the window is moved to the next position and the same process is repeated. If a face is detected, the face orientation is detected (S62). The result is output to a predetermined memory area (S63). By repeating the above processing in all images, the face position and face direction can be detected. By detecting the position of the face, the position where the person exists can be detected simultaneously. Finally, the processing of the entire image is confirmed (S64).

  Here, the details of the recognition unit 104 have been described using face detection as an example, but there are various other methods for acquiring information from an image. For example, if the discriminator is configured to detect the entire person instead of detecting the face, it is possible to realize human detection and the body orientation can be obtained in the same manner. Further, if the position of a person on the image is detected, the size (area on the image) can be obtained naturally. Further, it is possible to execute a person tracking process by obtaining a position where the detected area is moved over a plurality of frames (images) with continuous time.

  Moreover, if it is a vehicle, information on arbitrary images, such as a license plate and a driver | operator's face, can be acquired.

  Further, by taking a correspondence relationship between the position detected by the above-described processing and the monitoring area 501 and the camera image 502 described in FIG. 5, on the monitoring area 501 (the same applies to the monitoring area 205 in FIGS. 2 to 4). Position, movable direction (movement direction 206 in FIGS. 2 to 4), face direction (face direction 207 in FIGS. 2 to 4), and the like. By having this configuration, the correspondence between the monitoring area 501 and the camera image 502 indicates the direction in which the person facing right in the camera image 502 is facing in the monitoring area. That is, not only the position but also the direction can be acquired.

  FIG. 7 is an example of a data structure when the result obtained by the recognition unit 104 is stored in the recognition result 105. This data is composed of the object ID (D70), the position of the object in real space (D71), the area on the image (D72), the face direction (D73), the movement vector (D74), and other information (D75). Has been.

  Since the area (D72) differs depending on each camera that captures an object, the area (D72) is stored for each camera that captures an object, such as area / camera 1 (D76) and area / camera 2 (D77).

  Further, the movement vector 74 has information from the current time t to the information traced back for a certain period of time, and is stored as the position (t) (D79), and the movement direction (D78) is also stored from the information. The moving direction (D78) can be calculated from the average value of the information of the position (t) (D79). Similarly to the face direction, these pieces of information can also determine the moving direction on the monitoring area with respect to the direction on the camera image by determining the correspondence as shown in FIG.

  Other information (D75) can also be included in the data by adding processing of the recognition unit 104.

  Next, the display selection unit 106 shown in FIG. 1 will be described. First, the camera arrangement information 107 used in the display selection unit 106 will be described.

  The camera arrangement information 107 includes information indicating the positional relationship between the cameras and the relationship between the moving object and the camera image. The former can be obtained by obtaining the correspondence relationship between the camera and the monitoring area described above, and can be obtained by actually measuring in detail. The latter can be obtained by camera calibration.

  FIG. 8 shows an example of the positional relationship information of the camera in this embodiment.

  The camera positional relationship is an arbitrarily assigned camera ID (D80), a camera depression angle (D81), a camera horizontal angle (D82), a field angle (D83), and an installation position (D84). Each stores an angle and an absolute position. This makes it possible to define the direction in which the camera is facing and the video to be captured, and to grasp the positional relationship with each camera.

  When obtained by the camera calibration technique as described above, a perspective projection transformation matrix between the monitoring area 501 and the camera image 502 shown in FIG. 5 can be obtained, and this information may be obtained as the camera arrangement information 107 in some embodiments. It may be stored. Note that the positional relationship of the cameras can be added as information because the focal length, the rotation in the optical axis direction, and the size of the image sensor are also related.

  FIG. 9 is a bird's-eye view of the monitoring area 205 in the embodiment shown in FIGS.

  FIG. 9 shows cameras 200 to 202, an object 203, and a structure 204, as in FIGS. Here, it is assumed that data including the movement direction 206 and the face direction 207 is acquired through the recognition unit 104. In addition, the positional relationship of each camera is acquired from the camera arrangement information 107 in the same manner.

The cameras 200 to 202 are installed at angles of φ _{0 to} φ ₂ at respective positions in the Xw-Yw space of the monitoring area 205. The object 203 is moving in the direction of movement 206 (angle θv), and the face direction 207 of the object 203 is defined by θf.

  Using these pieces of information, the display selection unit 106 executes a process for determining an image to be displayed.

  FIG. 10 is an example of an information table obtained from the object 203, the camera position, etc. in the camera arrangement information 107. The camera evaluation items (D1000) of the information table are the distance (D1002) between the detected object and the camera, the area (D1003) on the camera image of the detected object, the face direction (D1004), and the moving direction (D1005). Composed. The evaluation value (D1001) is an evaluation value for each camera. Specifically, the camera evaluation value (D1000) is calculated based on the data of the object 203 obtained by the recognition unit 104 described above, the camera position, and the like, and is calculated and acquired for each camera, and the acquired value (D1001). ).

The distance to the camera (D1002) is obtained from the relationship between the position on the image detected by face detection or person detection and the monitoring area 205. The area (D1003) is also obtained from the similarly detected area. As the face direction (D1004), the face direction θf on the monitoring area 205 can be obtained from the face direction on the camera image, and can be calculated by the difference in angle with the directions φ _{0 to} φ ₂ of the cameras 200 to 202. it can. For example, the calculation formula for the camera 201 (camera 2 in FIG. 10) is shown in Equation 1.

[Equation 1]
Camera 2 face direction = (φ ₁ −θf) (1)
Here, as the face direction obtained by Equation 1 approaches 180 degrees, the imaging direction of the camera and the face direction face each other, that is, the face direction faces the camera direction. Strictly speaking, the angle in the vertical direction of the face with respect to the depression angle of the camera can be obtained, but in this embodiment, only the horizontal direction is described for simplicity.

  Further, since the moving direction (D1005) can also be obtained based on the same concept, description thereof is omitted here.

  Furthermore, in another embodiment, it is also possible to define a desired part such as a direction in which a specific part of an object is detected, for example, a direction having a possession such as a bag, a direction in which a part such as a hand is imaged. . When the object is a vehicle, an information table can be created based on the license plate, the driver's face, or the like. Moreover, it is good also as a direction which can observe a certain specific action (event). For example, a direction in which a person presses a button can capture an image, an action of picking up a product, and the like.

  This information table is stored in the camera arrangement information 107. As described above, the camera selection information and the recognition result acquired in advance are used together to determine the video selection in more detail. In addition, by feeding back the positional relationship with the camera and the result of the recognition process to the recognition process, it is possible to select a camera and a position on the video that are appropriate for the recognition process. Is possible.

  Next, the display selection unit 106 will be described with respect to a method for switching video display using the information table shown in FIG.

  FIG. 11 shows the priority setting of the camera evaluation items to be preferentially acquired among the camera evaluation items described above when using video information acquired from each camera in which the same object is detected in duplicate. It is an example of a setting screen 1100. These priorities are set, for example, via the input unit 108 in FIG. Respective priorities can be set on the screen, and in the present embodiment, priorities from 0 to 5 can be set. The case of 0 indicates that no information is used, and the case of 5 can use the corresponding information most preferentially. Here, a video to be output can be selected using only one piece of information, and a reference for video selection can be created by obtaining an evaluation value obtained by integrating several pieces of information. In the example shown in FIG. 11, the priorities are set as distance dp = 0, area sp = 1, face direction θfp = 3, and moving direction θvp = 1.

  Here, the distance of each camera is defined as d, the area is defined as s, the face direction is defined as θf, and the moving direction is defined as θv, and all the cameras are ranked with respect to each value from the camera arrangement information 107 shown in FIG. Taking the camera 1 as an example and defining the order of the camera 1 with respect to all cameras, the distance D1 = 3, the area S1 = 1, the face direction Θf1 = 1, and the moving direction Θv1 = 3. Since it represents the order in the camera, the minimum value is 1, and the maximum value is the number of cameras.

  A method for calculating the evaluation value for each camera for each arbitrary object based on the priority of each camera and the priority of the camera evaluation items is shown in Equation 2. The calculation of the evaluation value is performed in the display selection unit 106 shown in FIG.

[Equation 2]
Evaluation value (Camera 1)
= (D1 * dp + S1 * sp + Θf1 * θfp + Θv1 * θvp)
(2)
According to Equation 2, the display selection unit 106 can determine that the camera with the smallest evaluation value is a camera suitable for observing the object 203.

  For example, as a result of calculation based on the camera evaluation item information table and the camera evaluation item priority for each camera shown in FIGS. 10 and 11 with respect to the video shown in FIG. 4, the video shown in the camera image 401 of FIG. It is calculated that the video is most suitable for observing the object 203.

  Since the evaluation value for each camera is calculated for each object, a video image of the camera suitable for each moving object is defined by the evaluation value. In the case where there are a plurality of objects, there are coping methods such as controlling the person who is imaged the most, or performing the processing of the present invention only for the person selected on the input screen or the like.

  Next, the video display unit 109 shown in FIG. 1 will be described with reference to FIG. Here, it is assumed that the display selection unit 106 calculates a camera to be preferentially displayed among videos obtained from the video acquisition unit 103 and a video obtained from the camera. A part of the video example shown in FIG. 12 is the same as the example shown in FIG. 4, and the face direction of the moving person imaged by the video of the camera 2 is the camera direction, and the video is easy to observe. Yes.

  FIG. 12 shows a part of the monitoring screen of the monitoring system. Here, the video of each camera that images the monitoring area is displayed in the small window area 1200, and the video acquired through a video recording device, a matrix switcher, or the like is displayed. The video selected by the display selection unit 106 is further displayed in the large window area 1201. Although details about the setting of the multi-screen output and the like will not be described, the screen layout can be arbitrarily set depending on the number of cameras used.

  By superimposing the detection frame 1202 that detects the object on the large window area 1201 and outputting it, it is possible to draw the attention of the supervisor. Further, it is possible to output additional information 1203 on the screen using the result of face detection.

  The screen example shown in FIG. 12 also has a playback control unit 1204 that controls recorded video and the like, and can also be used for playback of recorded video. If the information table and the recognition result 105 shown in FIG. 10 are stored together with the recorded video, it is possible to realize a video that makes it easy to observe a person in the recorded video. The priority setting can be set from a setting button 1205 or the like.

  The screen example illustrated in FIG. 13 is a diagram illustrating an example in which the positional relationship between the camera and the moving object is visually easily displayed using the camera arrangement information 107. Camera images 1300 to 1302 are displayed close to the camera position on a screen on which data created by CG or the like for the monitoring area and the camera position is drawn. Each camera image can be displayed with its size changed by a constant coefficient according to the position on the screen.

  An observed object 1303 is drawn on the screen based on the position calculated by the recognition unit 104. This object 1303 is obtained by superimposing data created by extracting a moving person existing in the camera image calculated by the display selection unit 106 instead of a person model created by CG or the like on the screen. It is possible to observe the situation of 1303 at the same time. Further, by displaying additional information 1304 such as a face together, it is possible to observe in more detail.

  FIG. 14 is a diagram showing a case where a plurality of objects exist. In FIG. 14A, camera images 1400a to 1402a are displayed in the same manner as in FIG. An object 1403a and an object 1404a are drawn on the screen.

  For example, when it is desired to observe the image centered on the camera 2 image 1400a and the moving object 1404a, the viewpoint can be converted by an instruction from the user. When the user designates the camera image to be observed or the position of the moving object drawn on the screen from an input device such as a mouse, the viewpoint is switched to the viewpoint centered on the camera 1 image 1400b as shown in FIG. Although the positions of the moving object 1403b and the moving object 1404b are also changed by switching the viewpoint, images that are most easily observed are superimposed. In addition, the image size and the like of the video of each camera can be changed according to the distance from the viewpoint.

  In this way, depending on the importance or attention level of the video, the display size, the presentation with important information added, or the display method that presents the video in a format linked with the camera arrangement, It is possible to visually recognize the arrangement relationship in the monitoring area. As a result, it is possible to simultaneously grasp the importance of the video and the correspondence relationship between the monitoring areas, which leads to a reduction in the burden on the monitoring staff, and as a result, a more robust monitoring system can be provided. In addition, by giving priority to video display based on the priority order, it is possible to present and record a video suitable for observing an object, and to reproduce a video to be observed among recorded videos.

  The detection object to which the present invention can be applied includes a person as described above, and it is possible to perform face detection by recognition processing and select and present, for example, an image of a face captured from images from a plurality of cameras. is there. Moreover, even if it is not a person, this invention is applicable also to a vehicle, the bag which a person possesses, the screen of a personal computer, the door of a safe, etc. Determine the camera image suitable for monitoring the part you want to observe, such as monitoring the driver's face if it is a vehicle, or monitoring the face of a person carrying the bag or the bag itself. can do. In addition, you can select a camera that is suitable for observing parts that have moved or changed, such as when monitoring the screen of a computer whose orientation or screen display has changed, or when the safe door is opened, You can always monitor the computer screen or monitor the door only when the safe door is opened. The present invention can be applied not only to a movable object but also to a stationary object. For example, if you want to switch the monitoring area from the door side to the side when monitoring a fixed safe, use the configuration of the present invention to select a camera suitable for monitoring the side and switch the monitoring screen. be able to.

  FIG. 15 is an example of a configuration diagram in which the above-described embodiment is used for searching video data. Since the functional blocks shown in FIG. 15 are mostly the same as those shown in FIG. 1, only the portions related to this embodiment will be described.

  The video acquired by the video acquisition unit 103 is stored in the video data 1500. The search function in the monitoring system is one of means for acquiring the video data 1500 data. Video search conditions to be acquired from the video data 1500 are input to the search unit 1501 through the input unit 1502. There are various search conditions such as a time zone, a target camera, a specific person, and the like. The search unit 1501 here can also have a recognition function, like the recognition unit 104 in FIG. Similarly, the information acquired by this recognition function can be acquired in the same manner as the recognition result shown in FIG.

  The recognition result acquired here is used to prioritize the images for each camera in the display selection unit 106 as in the above-described embodiment. 110 can be displayed.

  Note that when the video data is stored in the video data 1500, the recognition result 105 may be stored at the same time. In this case, since it is not necessary to perform recognition processing in the search unit 1501, the search time can be shortened.

  Furthermore, based on the above-described embodiment, an embodiment used for improving recognition performance will be described with reference to FIG. Since some of the functional blocks are the same as those shown in FIG. 1, description thereof is omitted here. The video acquired by the video acquisition unit 103 is sent to the multiple camera recognition unit 1600. In this embodiment, since the images from a plurality of cameras may be processed at the same time, they are distinguished from the recognition unit 104. The result processed by the multi-camera recognition unit 1600 is sent to the video selection unit 1601. The video selection unit 1601 here has the same configuration as the display selection unit 106 described above. That is, after obtaining the recognition result, it is used to select a video suitable for recognition processing.

  The recognition results processed by the multiple camera recognition unit 1600 include those that can be expected to have high performance depending on the installation state of the cameras 100 to 102 and those that are not. Based on the result output by the multi-camera recognition unit 1600, the video selection unit 1601 calculates an evaluation value similar to the method shown in Equation 2, outputs a video suitable for recognition processing, and feeds back to the multi-camera recognition unit 1600 again. It is possible to improve recognition performance.

  For example, when face detection is taken as an example, which camera is most suitable for face detection can be determined by the recognition result (recognition rate). Furthermore, even on a single camera image, it is possible to calculate an area where a good result of face detection can be expected and an area where it is not. Therefore, in this embodiment, the multi-camera recognition unit 1600 can define a camera suitable for recognition and a region suitable for recognition in the camera image, and can be expected to realize a monitoring system with higher detection accuracy. .

  Further, when considering the accuracy of the person detection position, the camera image 400 in FIG. 4 has high vertical position accuracy, and the camera image 401 has high horizontal position accuracy. When determining the position of a moving object observed across a plurality of cameras on the monitoring area, it is also possible to select a result to be output based on such information. This makes it possible to perform highly accurate position detection.

100 to 102 Camera 103 Video acquisition unit 104 Recognition unit 105 Recognition result 106 Display selection unit 107 Camera arrangement information 108 Input unit 109 Video display unit 110 Display unit 200 to 202, 500 Camera 203, 1303, 1403, 1404 Object 204 Structure 205 , 501 Monitoring area 206 Moving direction 207 Face direction 300 to 302 Imaging area 400 to 402, 502, 1300 to 1302, 1400 to 1402 Camera image 504 Camera image position 505 Monitoring area position 1100 Priority setting screen 1200 Small window area 1201 Large window Area 1202 Detection frame 1203, 1304 Additional information 1204 Playback control unit 1205 Setting button 1500 Video data 1501 Search unit 1502 Input means 1600 Multiple camera recognition unit 1601 Video selection unit

Claims (13)

A plurality of cameras for imaging the surveillance area;
A recognition unit for detecting an object from images acquired by the plurality of cameras,
When an object is detected in a monitoring area that is imaged redundantly by the plurality of cameras, a recognition result that is a feature amount of the object is acquired by the recognition unit for each camera, and the recognition result and the recognition result A video monitoring system, comprising: a display selection unit that assigns a priority according to the priority to the video for each camera based on the priority. The video surveillance system according to claim 1,
The display selection unit calculates an evaluation value related to the object for each camera based on the recognition result and the priority of the recognition result, and prioritizes the video for each camera based on the evaluation value. A video surveillance system characterized by that. The video surveillance system according to claim 2,
Using camera installation information for each of the plurality of cameras to obtain camera arrangement information related to the object with respect to the plurality of cameras, and calculating the evaluation value using at least one of the camera arrangement information. A video surveillance system. In the video surveillance system according to claim 3,
The camera installation information includes information on camera position, depression angle, horizontal angle, angle of view, and rotation, and is calculated by acquiring a correspondence relationship between the plurality of cameras and the monitoring area. . In the video surveillance system according to any one of claims 1 to 4,
The recognition result is a moving direction, a size, and a predetermined area of the object. In the video surveillance system according to any one of claims 1 to 5,
Using the camera installation information for each of the plurality of cameras, obtain camera arrangement information including the distance of the object with respect to the plurality of cameras, the direction of the predetermined area of the object, and the moving direction, and out of the camera arrangement information A video surveillance system, wherein the evaluation value is calculated using at least one or more. In the video surveillance system according to any one of claims 1 to 6,
An image monitoring system comprising: a display unit configured to change an output form of an output image for each camera according to the priority order. The video monitoring system according to claim 7,
The output video output to the display unit displays the position of the monitoring area and the plurality of cameras, and combines the video acquired for each of the plurality of cameras with the output video for display. system. The video surveillance system according to claim 7 or 8,
The video monitoring system characterized in that the output video output to the display unit combines and outputs the moving object acquired from the recognition result and the predetermined area as additional information. In the video surveillance system according to any one of claims 7 to 9,
In the output video output to the display unit, the output video is reconstructed and output in an arrangement centered on the monitoring target by selecting the moving object or the video to be monitored. Monitoring system. In the video surveillance system according to any one of claims 1 to 10,
An image monitoring system for recording an output image for each camera on a recording medium according to the priority order. In the video surveillance system according to any one of claims 1 to 11,
The video monitoring system, wherein the plurality of cameras or the monitoring area of each camera to be processed by the recognition unit is selected according to the accuracy of the recognition result. A recognition unit that detects an object from images acquired by a plurality of cameras that capture an image within the monitoring area;
When an object is detected in a monitoring area that is imaged redundantly by the plurality of cameras, a recognition result that is a feature amount of the object is acquired by the recognition unit for each camera, and the recognition result and the recognition result A monitoring camera control device comprising: a display selection unit that assigns a priority according to the priority to the video for each camera based on the priority.

Images